I've just started reading up on Portfolio Optimization models and have come across the use of exposure bounds to mitigate the sensitivity of the optimized model solution, owing to parameter estimation errors. I'm just confused as to what exposure bounds are?
Before I answer your question, allow me to suggest that you clarify your question: let me propose that you edit your question to include the source, i.e. where exactly did you read this? I have found only very few resources mentioning exposure bounds.
Your question seems to imply that bounding the possible exposure of each [type of] asset or security in your portfolio (e.g. equities 20-60%; bonds 20-40%; etc.) will keep the sensitivity lower. Of course, if the ranges are smaller (bounded), the sensitivity could also only go so far. I am not sure which parameter estimation techniques you refer to - but again, if the ranges are smaller, parameter estimation could become easier (the search space is smaller) and less prone to errors.
Hope this helps. Again, feel free to provide more detail.